AI Made Curiosity Cheaper
Notes by SDG ·
In December 2024, my life looked pretty standard.
I was working full-time at Microsoft with a demanding schedule, two kids, school drop-offs, and an evening agenda dominated by dishes, laundry, and hunting for missing socks. I had no websites, no apps, no patents on the radar, and almost no presence on LinkedIn.
Fast forward to August 2026.
I now have three web apps and three mobile apps. One of them is already live in the App Store, and a few others are in different stages of testing and rollout. I write fairly regularly on LinkedIn and on my own site.
I recently joined the Advisory Board for the Strategic AI program at the University of San Francisco School of Management.
And I’m now working on filing three provisional patents around ideas that came out of things I was building.
When I look at that list, it sounds like I suddenly found a lot of free time. I didn’t.
I still have a full-time job. Two kids. School runs. Gymnastics. Taekwondo. Random household errands. Dishes. Laundry. The usual stuff.
The dishwasher, unfortunately, has not become agentic yet.
So what changed?
I think AI made curiosity cheaper.
Before AI, curiosity often stopped at curiosity
I’ve always had a tendency to ask questions.
Could this work differently?
Why does this product do it this way?
What if someone built something that did X?
But there was usually a fairly large gap between having the thought and actually doing something about it.
If the idea involved software, for example, the next thought was usually:
I don’t know how to build that.
And that was often the end of it.
Of course I could have learned.
People have been teaching themselves software development for decades.
But learning the architecture, frameworks, APIs, databases, authentication, deployment, App Store process and everything else before you can test one small idea is a pretty big hill to climb when you already have a job and a family.
AI changed that for me.
Now the conversation is more like:
“I wonder if this could work.”
Then:
“How would I build the simplest version?”
Then:
“Why is this broken?”
Then:
“Explain this error to me like I have never seen it before.”
Then, usually:
“No, that didn’t work either. Try again.”
And somehow the idea keeps moving.
A lot of what I built started with a question
AskNoodl started with me thinking about my own kids. If AI can answer almost any homework question, is that actually good for learning?
What if AI helped kids think instead of just giving them answers?
Could it ask questions back?
Could it explain something in different ways?
Could a kid teach the concept back to the AI?
Could you turn what they just learned into a small game?
That curiosity became AskNoodl.
Then I started thinking about family routines and all the little things families try to keep track of. That became Loopi.
More recently I started thinking about nutrition and fitness. Most calorie-tracking apps are very transactional.
You ate this.
You burned this.
Here is how much you have left.
But bodies don’t really work like a checking account.
Weight fluctuates. Wearables are noisy. Exercise calories are estimates. One high-calorie day shouldn’t necessarily cause the system to aggressively cut tomorrow’s target.
So I started asking different questions.
Could the system learn over time?
Could it look at weekly energy instead of obsessing over individual days?
Could it adjust targets only when there is enough evidence?
Could it avoid the common loop where reduced calories lead to lower weight, which leads to lower targets, which leads to even lower calories?
Those questions eventually became part of Kilo.
And while working through them, I found a few ideas that seemed genuinely different enough that I started exploring patents.
That was definitely not part of the original plan.
The biggest change is that I can follow the question much further
That, to me, is the interesting thing about AI.
It is not that AI suddenly knows everything.
It doesn’t.
I’ve had AI give me bad code, wrong answers, strange architectural recommendations and extremely confident explanations of things that turned out to be nonsense.
Building with AI still requires a lot of judgment.
But getting stuck feels different now.
Previously, reaching the edge of my knowledge often meant stopping.
Now it usually means asking another question.
Why did this fail?
What assumption am I making?
What are three other ways to do this?
What would an experienced engineer worry about here?
What am I not thinking about?
That ability to keep going is incredibly powerful.
One curiosity also tends to create another
This part surprised me.
Building apps made me curious about how AI systems actually work.
That led me deeper into models, agents and workflows.
Thinking about workflows made me think more about how companies are approaching AI.
And the more I watched companies approach AI through lists of use cases, the more I started wondering whether we were starting in the wrong place.
Organizations fundamentally make decisions and take actions.
So instead of only asking:
“Where can we use AI?”
Maybe leaders should also ask:
“What decisions are being made across the organization?”
“Which parts should humans continue to own?”
“Which parts can machines handle?”
“And where does a human-machine workflow make the decision better?”
I started writing about those ideas.
Writing forced me to make the thinking clearer.
That led to conversations with other people thinking about similar problems.
And eventually that contributed to me joining the advisory board for USF’s Strategic AI program.
That whole path was never planned.
It was just one question leading to another.
That is why I don’t think the real story is AI productivity
Productivity is certainly part of it.
I can do some things much faster now.
But I think the bigger change is this:
AI reduces the cost of trying.
Before, an idea might require weeks of learning before I could even figure out whether it was a bad idea.
Now I can often build a rough version, test it, learn from it and throw it away.
That completely changes the economics of curiosity.
You don’t need every idea to work.
You can explore ten.
Maybe eight go nowhere.
One teaches you something.
And one becomes interesting.
I also don’t think AI replaces expertise
If anything, building more has made me respect expertise more.
AI can help me build software.
That doesn’t make me a software engineer with twenty years of experience.
AI can help me think through nutrition research.
That doesn’t make me a nutrition scientist.
AI can help me understand patent concepts.
That doesn’t make me a patent attorney.
But something important has changed.
You no longer need to be an expert just to start exploring a question.
You can get far enough to learn whether the question is worth pursuing.
Then you bring in deeper expertise when it matters.
That opens up a lot of territory for people who are curious outside their formal job descriptions.
And somehow all of this fits around normal life
Most of these things didn’t happen during some six-month sabbatical.
They happened between work calls, after the kids went to bed, on weekends, during random thirty-minute windows and sometimes far too late at night.
AI didn’t give me more hours.
It made those small pockets of time much more useful.
Thirty minutes can be enough to test an idea.
An hour can be enough to build a feature.
A random question can become a blog post.
A blog post can become a conversation.
A conversation can turn into something you never expected.
That is basically what happened to me between December 2024 and August 2026.
I didn’t suddenly become dramatically more ambitious.
I think I just became much more willing to follow my curiosity.
Because now, when I wonder:
“Can I build this?”
I don’t automatically assume the answer is no.
I usually open an AI tool and find out.
Historically, expertise acted as a gatekeeper. AI lowers that gate just enough for curiosity to walk through.
Once you're on the field, reality still takes over—users only care if the app works, readers only care if the insight is real, and patent examiners only care if the idea is novel. AI doesn't guarantee your success, but it gets you into the game.
I still don't know which of my experiments will matter five years from now. But AI made me far more willing to follow the questions—and that’s where all the fun is.